Automette reads invoices, orders, statements and forms into the fields you need, has a person check anything that fails your rules, and sends the data to the apps you already use.
Brightline Paper Co
14 Mill Lane, Leeds
Invoice INV-2291
Date 02/10/2026
PO number PO-7781
A4 paper × 1045.00
Toner × 276.00
Total121.00
Extracted data
Click a value to see where it came from.
It extracts data from PDFs, scans and photos, Word files, spreadsheets and emails, with no template to set up for each layout or each supplier. Scans and photos are read with OCR, and tables and line items are read even when they run across pages. Some of the documents it is used for:
Data extraction runs as an agent, so each document goes through the whole process: it comes in, is sorted, read, checked, and sent where it needs to go.
1Ingest any format
PDFs, scans, photos, Excel, Word and emails, from an inbox, an upload, a form or the API.
2Classify and split
A scanned packet is split into its documents, and each is sorted by type.
3Extract the data
Fields, tables and line items, each linked to where it was found on the page.
4Review exceptions
Values are checked against your rules. What fails goes to a person, with the reason.
5Deliver
To Google Sheets, Airtable, a spreadsheet, a webhook or your software through the API.
Each value is linked to the place it was read from, so anyone checking it can click it and see it on the page. Nothing that fails a check is sent on without a person seeing it.
Review: invoice INV-3317, Harbor Office Supply
Corrected by Priya, then approved.
Suggested rule
Harbor Office Supply writes dates day first: read 10/03 as 10 March.
Once a document is read and checked, the agent sends the data where your team works with it, and can use it for the next step of the process.
Intelligent document processing (IDP) is reading business documents into structured data with OCR and AI, checking it, and passing it on, instead of typing it in by hand. In Automette it runs as an agent: documents come in, are sorted and read, checked against your rules, and the data is delivered to your apps.
No. You say which fields you want, in plain words or as a list, and the same fields come back whatever the document looks like.
Yes. Scans and photos are read with OCR, and each value is linked to the place on the page it came from, so it can be checked.
Yes. Line items and other tables come back as rows, even when they run across several pages.
Values that fail your rules go to a person, who corrects them before anything is sent on. A correction can become a rule, so the agent gets it right next time.
Set up an extraction agent, send it a document, and see the fields it reads and where each one came from. Free to start.